Modelling the Compressive Strength of Bituminous Sand Concrete

Global demand for river sand now outstrips natural replenishment, driving the search for sustainable alternatives. This study characterizes bituminous sand from bitumen deposits in Ondo State, Nigeria, as a substitute for river sand in concrete and develops regression and machine-learning frameworks to predict compressive strength. Aggregate characterization recorded effective sizes (D 10 ) of 0.123 mm (river sand) and 0.099 mm (bituminous sand), with uniformity coefficients of 3.959 and 4.495, indicating well-graded materials. Moisture contents averaged 72%, 60% and 8% for river sand, bituminous sand and granite; silt contents averaged 2.46% and 4.98%, against an 8% permissible limit. Specific gravities were 2.64, 2.60 and 2.70, bulk densities 1780, 1717 and 1677 kg/m 3 , and water absorption 0.91%, 0.70% and 0.74% respectively. Using a grade M25 mix of ratio 1: 1: 2 (characteristic strength 25 N/mm 2 at 28 days, target 38 N/mm 2 ), BSC showed higher workability (slump 55 mm vs 30 mm, compacting factor 0.49 vs 0.46) and strength gain with curing age, from 9.38 N/mm 2 at 7 days to 15.35 N/mm 2 at 150 days, against 15.17-26.93 N/mm 2 for the control (36-43% differences). BSC density (2213 kg/m 3 , 13.41 N/mm 2 ) was within 10% of the control (2462 kg/m 3 , 22.32 N/mm 2 ), classifying BSC as normal-weight concrete, and water absorption stayed within the 10-15% band (averages 11.37% and 11.30%). Multi-model regression: the exponential curing-age model fitted the control best (R 2 = 0.8692) and the logarithmic model fitted BSC best (R 2 = 0.8008), with adequacy confirmed by F-statistics of 187.68 and 757.64 and a t-test (2.14 < 2.57). A five-stage machine-learning framework is then specified, benchmarked against ANN achieving R 2 = 0.894 with the lowest RMSE (5.192 MPa) on a 1234-record dataset. The results position bituminous sand as a technically viable, sustainable alternative fine aggregate with a data-driven strength-prediction route.